Keep building with your AI tool

You don’t have to stop using the tool you built your app with. Your AI tool keeps working on the code in GitHub, and ScaleBop deploys the changes you choose to your AWS account.

How changes reach production

1. Your AI tool commits to your repository’s default branch (usually main), often once per prompt.
2. Those commits don’t go live on their own. Your app on AWS keeps running the last version you shipped.
3. When you’re happy with the changes, open your project in ScaleBop and select Ship changes. ScaleBop shows how many changes are waiting.
4. ScaleBop moves the production branch up to your default branch, and GitHub Actions deploys it.

Your AI tool’s own preview stays your place to try things out. ScaleBop production is what your users see.

Auto-ship

If you’d rather every commit go live straight away, turn on Auto-ship on your project hub. Each push to your default branch then ships automatically. Half-finished changes go live too, so most builders leave it off. Auto-ship still holds back changes your current setup can’t run (see When your app changes shape).

Files your AI tool shouldn’t change

ScaleBop adds these to your repository:

• infra/: the AWS infrastructure for your app
• .github/workflows/deploy.yml: the workflow that deploys it
• scalebop/: ScaleBop’s record of the files it generated

If your AI tool edits or deletes them, the next deployment can fail. If a change needs a different setup, update it in ScaleBop instead.

Telling your AI tool

ScaleBop adds rules for AI coding tools in the same pull request:

• a ScaleBop section in AGENTS.md and CLAUDE.md. ScaleBop only changes what’s between its markers; the rest of each file stays yours.
• .cursor/rules/scalebop.mdc for Cursor

Tools you use in the browser, such as Lovable and Bolt, take instructions you set inside the tool instead. Copy the rules from Deployment files → Rules for your AI tool on your project hub and paste them into your project’s knowledge or custom instructions there.

Your build and lint

infra/ is a separate project with its own dependencies. If your root tsconfig.json or ESLint config would include it, ScaleBop adds infra to its exclude or ignore list in the same pull request, so your app’s own build and lint skip it. Nothing else in those files changes.

If ScaleBop can’t safely make that change (for example, your tsconfig.json extends another local file), it tells you the exact line to add before it generates anything. Your AI tool can make the change, then run the analysis again.

If files were changed or deleted

Deployment files on your project hub checks the files on your default branch. If any were changed or deleted, select Restore files to open a pull request that puts them back. A restore doesn’t count as a regeneration.

When your app changes shape

Some prompts change more than code. Your AI tool might turn a static site into an app that needs a server, move the build output to a different folder, or remove the Dockerfile your app deploys with. The setup ScaleBop generated can’t run that version.

ScaleBop checks each push to your default branch for changes like these:

• Changes the current setup can’t run pause Ship changes, and the deploy workflow stops before it builds anything. Your app on AWS keeps running the last version you shipped. Select Update setup on your project hub: ScaleBop updates the setup and opens a pull request with the new files. Merge it, then ship. Update setup doesn’t use a regeneration.
• Other changes show as notes on the Ship changes card and don’t stop anything. Examples: a new environment variable your app uses, or a second app in your repository.

If ScaleBop can’t be reached, the deploy workflow skips the check and deploys anyway.

Environment variables

When your app needs a new setting, such as an API key, add it under Configuration on your project hub as well as in your code. Adding it only in your AI tool doesn’t make it available to the app on AWS.

When a deployment fails

Each failed deployment on your project hub has a Copy fix prompt button. The prompt tells your AI tool:

• what failed, and the error output (with secrets removed)
• which files the error points to
• the likely cause and the change needed
• which files not to touch

Paste it into your AI tool, let it make the fix, then ship again. For causes that aren’t in your app code, see Troubleshooting.